In the past decades, a large volume of research work has been done on mechanism of rill formation by domestic and foreign scholars. Soil detachment as the main source of rill sediment is also a hot issue of the research on rill erosion. Being an important parameter of rill erosion models, like WEPP, soil detachment rate has also been a focal point in the research on rill erosion. In this article, experimental and analytic methods were proposed and tested for use in calculating rill detachment rate, with a view to defining suitable principles and methods for computing rill detachment ratesin the purplish soil region and enriching the basic research on rill detachment rate in purple soil with theories. The research adopted the in-lab soil flume scouring experiment, using flumes 12m in length each. The flumes were packed with purplish soil to form a sheet of plow layer over a simulated purplish subsoil layer low in permeability. The experiment was designed to have five slope gradients for the flumes (5°, 10°, 15°, 20° and 25°) and three water flow rates (2 Lmin^-1, 4 Lmin^-1 and 8 Lmin^-1). Water flow rates in the experiment were set in the light of the critical intensities of the rainfalls that produce rill erosion in sloping croplands on purple soil. Sediment was collected in a sampling bucket placed at the rill outlet. After a certain period of water scouring, rill erosion occurred, and then the soil flume was adjusted from sloping to horizontal. Plastic film was folded into multiple layers to form eleven thin, waterproof baffle plates the same in width as the rill. The baffle plates were inserted at 0.5 m, 1 m, 2 m, 3 m, 4 m, 5 m, 6 m, 7 m, 8 m, 10 m, and 12m from the rill entrance, thus dividing the erosion rill into the 11 sections: 0～0.5 m, 0.5～1 m, 1～2 m, 2～3 m, 3～4 m, 4～5 m, 5～6 m, 6～7 m, 7～8 m, 8～10m and 10～12 m. The baffle plates were inserted deep enough to prevent water from flowing between the rill sections. Each section of the rill was then filled with water and by calculating volume of the water in the rill, the soil erosion volume was worked out section; then quantity of the soil erosion was figured out from bulk density of the packed soil. Rill erosion was found to be quitelow on a slope 5° in gradient with a flow rate of 8 Lmin^-1, thus excluding the need of running the test with flow rates being lower than 8 Lmin^-1,that is, 2 Lmin^-1or 4 Lmin^-1, on the slope. Similarly, the test with the flow rate of 2 Lmin-1on a slope 10° in gradient, because erosion was already very low with flow rate at 4 Lmin^-1. All the other combinations of gradients and flow rates were tested, forming a total of twelve sets of experimental conditions. Each test had three replicates, thus making up a total of 36 separate trials. A series of process data were obtained through the experiment, and then rill detachment rates were calculated relative to hydraulic condition. On such a basis the relationships of rill detachment rate with rill length distribution process and sediment concentration were simulated. Results of the simulation show that rill detachment rate decreased exponentially with growing rill length, and linearly with rising sediment concentration. In the flumes higher in slope gradient and in flow rate, too, the effect of the simulation was better. Rill detachment rates were calculated with the analytic method and compared with those obtained through the experiment in their relationships with rill length and sediment concentration. It was found that the estimation through the experiment agreed well with the calculation using the analytic method, which indicates that the experimental method used in this research is effective and feasible.